Yuya Cui , Degan Zhang , Honghu Li , Hao Qiang , Haitao Zhao
{"title":"基于多智能体深度强化学习的车辆边缘计算协同任务卸载策略","authors":"Yuya Cui , Degan Zhang , Honghu Li , Hao Qiang , Haitao Zhao","doi":"10.1016/j.future.2025.107950","DOIUrl":null,"url":null,"abstract":"<div><div>Vehicular Edge Computing (VEC) and Vehicle- to-Vehicle (V2V) offloading can significantly reduce in-vehicle task latency. This paper investigates a cooperative task offloading strategy in VEC, where latency-sensitive and computation -intensive tasks can be offloaded to Road Side Units (RSUs) using 5 G connectivity. Additionally, these tasks can be shared among nearby vehicles through V2V links. Joint VEC and V2V cooperative offloading can not only minimizes task execution delays but also prevents network congestion. When vehicles are in motion, dynamic migration of computation tasks is necessary to maintain service continuity. We propose a two-phase distributed task offloading and migration strategy for multiple vehicles. In the first phase, vehicles select the optimal service vehicle based on inter-vehicle link quality and offloading willingness. In the second phase, to minimize system cost, we introduce a multi-agent reinforcement learning (MARL) based distributed task offloading and migration strategy. This strategy allows vehicles to choose the optimal edge node in a dynamic environment without fully offloading information. Moreover, we implement a counterfactual multi- agent (COMA) reinforcement learning approach to address the inefficiency caused by the credit allocation problem in multi-agent systems. Extensive evaluations demonstrate that the algorithm proposed in this paper perform better in terms of average system latency and overall task completion rate. Compared with related scheme, the proposed method can reduce latency by up to 54 % and improve task completion rate by up to 15 % in different scenarios.</div></div>","PeriodicalId":55132,"journal":{"name":"Future Generation Computer Systems-The International Journal of Escience","volume":"174 ","pages":"Article 107950"},"PeriodicalIF":6.2000,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Cooperative Task Offloading Strategy for Vehicular Edge Computing Based on Multi-Agent Deep Reinforcement Learning\",\"authors\":\"Yuya Cui , Degan Zhang , Honghu Li , Hao Qiang , Haitao Zhao\",\"doi\":\"10.1016/j.future.2025.107950\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><div>Vehicular Edge Computing (VEC) and Vehicle- to-Vehicle (V2V) offloading can significantly reduce in-vehicle task latency. This paper investigates a cooperative task offloading strategy in VEC, where latency-sensitive and computation -intensive tasks can be offloaded to Road Side Units (RSUs) using 5 G connectivity. Additionally, these tasks can be shared among nearby vehicles through V2V links. Joint VEC and V2V cooperative offloading can not only minimizes task execution delays but also prevents network congestion. When vehicles are in motion, dynamic migration of computation tasks is necessary to maintain service continuity. We propose a two-phase distributed task offloading and migration strategy for multiple vehicles. In the first phase, vehicles select the optimal service vehicle based on inter-vehicle link quality and offloading willingness. In the second phase, to minimize system cost, we introduce a multi-agent reinforcement learning (MARL) based distributed task offloading and migration strategy. This strategy allows vehicles to choose the optimal edge node in a dynamic environment without fully offloading information. Moreover, we implement a counterfactual multi- agent (COMA) reinforcement learning approach to address the inefficiency caused by the credit allocation problem in multi-agent systems. Extensive evaluations demonstrate that the algorithm proposed in this paper perform better in terms of average system latency and overall task completion rate. Compared with related scheme, the proposed method can reduce latency by up to 54 % and improve task completion rate by up to 15 % in different scenarios.</div></div>\",\"PeriodicalId\":55132,\"journal\":{\"name\":\"Future Generation Computer Systems-The International Journal of Escience\",\"volume\":\"174 \",\"pages\":\"Article 107950\"},\"PeriodicalIF\":6.2000,\"publicationDate\":\"2025-06-06\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Future Generation Computer Systems-The International Journal of Escience\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S0167739X25002456\",\"RegionNum\":2,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"COMPUTER SCIENCE, THEORY & METHODS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Future Generation Computer Systems-The International Journal of Escience","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0167739X25002456","RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, THEORY & METHODS","Score":null,"Total":0}
Cooperative Task Offloading Strategy for Vehicular Edge Computing Based on Multi-Agent Deep Reinforcement Learning
Vehicular Edge Computing (VEC) and Vehicle- to-Vehicle (V2V) offloading can significantly reduce in-vehicle task latency. This paper investigates a cooperative task offloading strategy in VEC, where latency-sensitive and computation -intensive tasks can be offloaded to Road Side Units (RSUs) using 5 G connectivity. Additionally, these tasks can be shared among nearby vehicles through V2V links. Joint VEC and V2V cooperative offloading can not only minimizes task execution delays but also prevents network congestion. When vehicles are in motion, dynamic migration of computation tasks is necessary to maintain service continuity. We propose a two-phase distributed task offloading and migration strategy for multiple vehicles. In the first phase, vehicles select the optimal service vehicle based on inter-vehicle link quality and offloading willingness. In the second phase, to minimize system cost, we introduce a multi-agent reinforcement learning (MARL) based distributed task offloading and migration strategy. This strategy allows vehicles to choose the optimal edge node in a dynamic environment without fully offloading information. Moreover, we implement a counterfactual multi- agent (COMA) reinforcement learning approach to address the inefficiency caused by the credit allocation problem in multi-agent systems. Extensive evaluations demonstrate that the algorithm proposed in this paper perform better in terms of average system latency and overall task completion rate. Compared with related scheme, the proposed method can reduce latency by up to 54 % and improve task completion rate by up to 15 % in different scenarios.
期刊介绍:
Computing infrastructures and systems are constantly evolving, resulting in increasingly complex and collaborative scientific applications. To cope with these advancements, there is a growing need for collaborative tools that can effectively map, control, and execute these applications.
Furthermore, with the explosion of Big Data, there is a requirement for innovative methods and infrastructures to collect, analyze, and derive meaningful insights from the vast amount of data generated. This necessitates the integration of computational and storage capabilities, databases, sensors, and human collaboration.
Future Generation Computer Systems aims to pioneer advancements in distributed systems, collaborative environments, high-performance computing, and Big Data analytics. It strives to stay at the forefront of developments in grids, clouds, and the Internet of Things (IoT) to effectively address the challenges posed by these wide-area, fully distributed sensing and computing systems.